HW9-report

Perform exploratory analyses

library(downloader)
download(url = "https://raw.githubusercontent.com/jennybc/gapminder/master/inst/gapminder.tsv", destfile = "gapminder.tsv")

#Bring in the data as data frame
library(readr)
gapminder <- read.delim("gapminder2.tsv")
str(gapminder)

#Descriptive plots
#remove the oc
library(ggplot2)
library(dplyr)
freq <- as.data.frame(table(gapminder$continent))
names(freq) <- c("continent","obs")
kable(freq,format = "markdown")

barplot(table(gapminder$continent))

p1 <- ggplot(gapminder,aes(x=lifeExp)) + geom_histogram()+ggtitle("lifeExp distribution")
p2 <- ggplot(gapminder,aes(x=gdpPercap)) + geom_histogram()+ggtitle("gdpPercap distribution")
p3 <- ggplot(gapminder,aes(x=continent, y=gdpPercap)) + geom_boxplot(aes(fill=continent)) + 
    scale_y_log10() + ggtitle("log(gdpPercap) distribution in different continents")
p4 <- ggplot(gapminder,aes(x=year, y=gdpPercap)) + 
    geom_point(aes(color = country)) + 
    geom_smooth(lwd = 2, se= FALSE, method = "lm") + 
    scale_y_log10() + facet_grid(~continent) +
    ggtitle("log(gdpPercap) growth in different continent")

ggsave(file="lifeExp_dist.png",plot=p1)
ggsave(file="gdpPercap_dist.png",plot=p2)
ggsave(file="gdpPercap_dist_continent.png",plot=p3)
ggsave(file="gdpPercap_growth_continent.png",plot=p4)

# Reorder the continents based on life expectancy according to the cross-country 
# difference of lifeExp in each continent 2007. 

gap <- gapminder %>% 
    filter(year==2007) %>% 
    group_by(continent) %>% 
    summarise(diff = max(lifeExp)-min(lifeExp))

gap <- full_join(gapminder,gap)
gapminder <- gap %>% 
    mutate(continent=reorder(continent,diff))
levels(gapminder$continent)
# sort the data according to continent, country, and year variable
sort <- gapminder[order(gapminder$continent,gapminder$country, gapminder$year),]

write_tsv(sort,"gap_sort.tsv")

I reorder the continent factor according to the biggest difference between countries in the same continent in year 2007. And add the variable diff into gapminder data in order to quickly reorder continent when reloading the file. Descriptive plots:

Perform statistical analyses

The “worst” and “best” countries are picked according to the value of intercept, which is seen as the “default” lifeExp.

# import the data from download.R
library(readr)
gapminder <- read.delim("gap_sort.tsv") #I want continents treated as factors
str(gapminder)
#make sure new continent order is still in force
gapminder <- gapminder %>% 
    mutate(continent=reorder(continent,diff))
levels(gapminder$continent)

library(plyr)
library(dplyr)
# linear regression of life expectancy on year
reg <- ddply(gapminder, ~ country + continent,fn <- function(dat, offset = 1952) {
    the_fit <- lm(lifeExp ~ I(year - offset), dat)
    sd <- var(residuals(the_fit), na.rm = TRUE)
    setNames(c(coef(the_fit),sd), c("intercept", "slope","sd"))
})

reg <- reg[order(reg$continent,reg$country),]
#write results to file
write_tsv(reg,"reg_life_year.tsv")

# find the 3 or 4 “worst” and “best” countries for each continent
rank_int <- reg %>% 
    group_by(continent) %>% 
    filter(min_rank(intercept)<5 | min_rank(desc(intercept))<5)

write_tsv(rank_int,"reg_worst_best.tsv")

Generate figures

library(dplyr)
library(readr)
library(ggplot2)
gapminder <- read_tsv("gap_sort.tsv")
rank_int <- read_tsv("reg_worst_best.tsv")
pick <- semi_join(gapminder,rank_int,by="country")
pick$continent <- as.factor(pick$continent)
pick$country <- as.factor((pick$country))
str(pick)


for (i in levels(pick$continent)){
    ggplot(pick %>% filter(continent==i),aes(x = year,y = lifeExp)) +
        geom_smooth(se = TRUE, method = "lm") + 
        geom_point() + facet_wrap(~country)+ggtitle(paste("lifeExp changes in Best and Worst countries in",i,sep=" "))
    k <- paste("lifeExp_growth_",i,".png",sep="")
    ggsave(k)
}

The final analysis graphs of for each continent.